Re: [PATCH] Fix pg_rewind false positives caused by shutdown-only WAL

On Tue, Sep 30, 2025 at 1:24 PM Srinath Reddy Sadipiralla
<srinath2133(at)gmail(dot)com> wrote:
> Can you please once confirm this, did you mean that this is not even an actual problem to fix or only this patch's logic which I provided does not make sense?, because i am trying out come up with another patch based on your inputs regarding considering controlfile changes , ignoring RUNNING_XACTS records, and to use XLogRecGetRmid test.

Well, the patch's idea is that we can ignore certain WAL records when
deciding whether pg_rewind is needed. But I do not think we can do
that, because (1) those WAL records might do important things like
update the control file and (2) the server will not be OK with
ignoring those WAL records even if pg_rewind decides that they are not
important. If you have a plan for working around those two issues,
please say what your plan is. I don't personally see how it would be
possible to work around those issues, but of course somebody else
might have a good idea that has not occurred to me.

--
Robert Haas
EDB: http://www.enterprisedb.com